AI-facilitated individualized nutritional guidance with ABC-structured prompts improves short-term dietary intake adequacy and patient-reported symptoms after gastrointestinal cancer surgery: a randomized controlled trial

BACKGROUND: Postoperative nutritional management is critical for gastrointestinal cancer surgery patients, yet discharge advice is often standardized and lacks individualization. Whether an AI-facilitated individualized and quantified strategy improves short-term dietary intake compared with standard generic advice remains unclear; this trial does not isolate the independent effect of AI. OBJECTIVE: To determine whether AI-facilitated individualized nutritional guidance using structured prompts improves short-term dietary intake adequacy and patient-reported symptoms versus standard discharge advice. METHODS: In this single-center, randomized controlled trial, 90 patients undergoing laparoscopic gastric or colorectal cancer surgery were assigned 1:1 to AI-facilitated individualized guidance generated by DeepSeek v3.2 using an ABC-structured prompt plus clinician review or standard discharge dietary advice (standard care group). The intervention combined AI, individualization, quantified targets, written guidance, and clinician review; the trial cannot isolate AI alone. Primary outcome was energy intake adequacy (≥ 90% of target) at 30 days post-discharge. Secondary outcomes included protein intake adequacy, nutritional status, patient-reported outcomes, process measures (expert-rated quality scores, provider time), and 30-day complications. Outcome assessors were blinded to group assignment. RESULTS: Baseline characteristics were well-balanced between groups. The AI-guided group achieved significantly higher energy adequacy (66.7% vs. 40.0%, OR 3.0, 95% CI 1.3-7.1, P = 0.01) and protein adequacy (62.2% vs. 35.6%, OR 2.9, 95% CI 1.2-6.8, P = 0.01) at 30 days. Expert-rated quality scores were superior in the AI-guided group for completeness (4.6 vs. 3.2), individualization (4.8 vs. 2.8), and actionability (4.5 vs. 3.5) (all P < 0.001), while provider time per patient was reduced (8.2 vs. 12.5 min, P < 0.001). Patient-reported outcomes showed significant improvements in fatigue, nausea/vomiting, appetite loss, and gastrointestinal symptoms (GSRS total score 2.1 vs. 2.5, P = 0.03) in the AI-guided group. No significant differences were observed in weight change, serum nutritional markers, or complication rates. CONCLUSION: AI-facilitated individualized nutritional guidance improved short-term dietary intake adequacy and patient-reported symptoms and reduced provider time compared with standard generic advice. These findings do not show that AI itself is superior to similarly individualized non-AI counseling and do not establish longer-term nutritional or clinical benefit. Clinical supervision remains essential.